Lu et al. 2003. Table 1 Supplement. Accuracies and representative coverage of current sub-cellular localization predictors.

Year System Accuracy Coverage Technique
2003 PSORT-B (Gardy et al.) 74.8% 1443 Gram negative combination of other systems
2002 LOCkey (Nair and Rost) 87% (test set)
8000 unverified predictions
37% of 3146 test proteins from Swiss-Prot keyword-vector based classifier
2001 SubLoc (Hua and Sun) 91.4% (prokaryotic)
79.4% (eukaryotic)
997 prokaryotic in 3 classes
2427 eukaryotic proteins in 4 classes
subset of SWISSPROT with verified localizations
SVM on amino acid compositions
2000 TargetP (Emanuelsson et al.) 85% (plant)
90% (non-plant)
940 plant (redundancy-reduced Arabidopsis chr 2,4)
2738 non-plant (Ensembl human proteins)
based on N-terminal amino acid sequence
1997
SignalP (Nielsen et al)
68.0% in Human,
70.2% in Eukaryote,
83.7% in E.coli,
79.3% in Gram-,
67.9% in Gram+
667 human, 1831 Eukaryotes,
224 E.coli, 452 Gram- and 205 Gram+
signal peptide detection
1997
ProtLock (Cedano et al.)
75% (test set)
200 proteins from Swiss-Prot amino acid composition
1997 NNPSL (Reinhardt and Hubbard)
66% (eukaryotes excluding plants)
81% prokaryotes
3420 proteins from Swiss-Prot amino acid composition
1997 PSORT (II) (Horton and Nakai)
86% (E.coli)
60% (yeast)
336 E.coli in 8 classes
1484 yeast proteins in 10 classes
k-nearest neighbor
1996 PSORT (II) (Horton and Nakai) 86% (E.coli)
60% (yeast)
336 E.coli in 8 classes
1484 yeast proteins in 10 classes
probabilistic classification tree
1992
PSORT (I)
66% of the training data
59% of the testing data
1484 Yeast sequences
expert system

References